Fault Aware Analog Model (faam)

1. A fault aware analog model (FAAM) system, comprising a FAAM builder module comprising a model construction module configured to: receive a reference dataset associated with a circuit block, the reference dataset comprising a set of data values that defines an input to output relationship of the circuit block for both in spec and out of spec operation of the circuit block, wherein the set of data values of the reference dataset comprises a set of input values that covers in spec and out of spec range of input values for the circuit block, and a corresponding set of output values measured using a ground truth representation of the circuit block; and generate a FAAM comprising a behavioral model of the circuit block, based on the reference dataset, wherein the FAAM is configured to approximate the input to output relationship of the circuit block that is defined by the set of data values in the reference dataset.

2. The FAAM system of claim 1 , wherein, in order to generate the FAAM, the model construction module is configured to: select a parametric model comprising a plurality of model parameters; split the reference dataset into at least a training dataset comprising a set of training data values and a test dataset comprising a set of test data values, wherein both the set of training data values and the set of test data values comprises a respective set of input values and a corresponding respective set of output values associated therewith; and train the selected parametric model based on the training dataset, to form an optimized parametric model comprising a plurality of optimized model parameters that are configured to approximate the input to output relationship of the circuit block that is defined by the set of training data values in the training dataset.

3. The FAAM system of claim 2 , wherein the selected parametric model is trained by modifying the plurality of model parameters to form the plurality of optimized model parameters, based on the set of training data values in the training dataset, in accordance with a predefined goodness of fit function.

4. The FAAM system of claim 2 , wherein the model construction module is further configured to validate the optimized parametric model, based on an approximation error associated with output values provided by the optimized parametric model, when the optimized parametric model is provided with the set of input values in the test dataset, in order to form the FAAM, wherein the approximation error comprises an error in the output values provided by the optimized parametric model with respect to corresponding output values in the test dataset.

5. The FAAM system of claim 4 , wherein, when the approximation error is lesser than or equal to a predefined validation threshold, the optimized parametric model is identified as satisfactory, thereby forming the FAAM.

6. The FAAM system of claim 5 , wherein, when the approximation error is greater than the predefined validation threshold, the optimized parametric model is identified as unsatisfactory, and the model construction module is configured to iteratively generate one or more updated optimized parametric models based on updating the reference dataset or based on updating the parametric model, until the approximation error associated with an updated optimized parametric model is lesser than or equal to the predefined validation threshold, to form the FAAM.

7. The FAAM system of claim 2 , wherein, when the selected parametric model comprises a static model, the model construction module is further configured to reshuffle the reference dataset, prior to splitting the reference dataset into the training dataset and the test dataset.

8. The FAAM system of claim 1 , wherein, in order to generate the FAAM, the model construction module is configured to utilize a nominal model of the circuit block that defines the input to output relationship of the circuit block using a mathematical relation between a set of input parameters and output parameters of the circuit block, and append the nominal model with one or more lookup tables (LUTs) derived based on the reference dataset, wherein each LUT of the one or more LUTs includes information of a variation of one or more output parameters of the nominal model for one or more input parameters of the nominal model.

9. The FAAM system of claim 1 , further comprising a data generation module configured to generate the reference dataset associated with the circuit block.

10. The FAAM system of claim 9 , wherein, prior to the generation of the reference dataset, the data generation module is configured to: identify input interfaces and output interfaces that corresponds to a type of the set of input values and the set of output values, respectively, to be included in the reference dataset; and discretize an input space comprising a range of input values that includes the in spec and out of spec range of input values for the circuit block to form the set of input values to be included within the reference dataset.

11. The FAAM system of claim 1 , wherein the FAAM builder module further comprises a model deployment module configured to convert the FAAM into a deployable format, prior to providing the FAAM to a simulator environment.

12. The FAAM system of claim 11 , wherein converting the FAAM into the deployable format comprises serializing the FAAM into a computer readable format, in order to store the FAAM to a non-transitory computer readable storage medium.

13. The FAAM system of claim 11 , wherein converting the FAAM into the deployable format comprises implementing the FAAM in a hardware description language (HDL).

14. The FAAM system of claim 1 , further comprising a simulator module that comprises a circuit block simulator module configured to receive the FAAM generated at the model construction module or a deployable version of the FAAM, in order to perform simulations of the circuit block using the FAAM.

15. The FAAM system of claim 14 , wherein the circuit block simulator module is further configured to receive the ground truth representation of the circuit block, and perform simulations of the FAAM and the ground truth representation using a same set of input values.

16. The FAAM system of claim 15 , wherein the simulator module further comprises a test module configured to receive output values associated with the simulations of both the ground truth representation and the FAAM, and compare the output values of both the ground truth representation and the FAAM, in order to verify a performance of the FAAM in a simulator environment.

17. The FAAM system of claim 14 , wherein the simulator module further comprises an application program interface (API) module configured to couple the circuit block simulator module to one or more machine learning libraries, in order to enable the circuit block simulator module to process and simulate the received FAAM.

18. A simulator module, comprising: a circuit block simulator module configured to simulate a fault aware analog model (FAAM) circuit block comprising a FAAM of a first circuit block, wherein the FAAM comprises a behavioral model that is generated based on a reference dataset comprising a set of data values that comprises a set of input values that covers in spec and out of spec range of input values for the first circuit block, and a corresponding set of output values measured using a ground truth representation of the first circuit block, and wherein the FAAM is configured to approximate an input to output relationship of the first circuit block for both in spec and out of spec operation of the first circuit block as defined by the set of data values in the reference dataset.

19. The simulator module of claim 18 , wherein the circuit block simulator module is further configured to simulate a ground truth circuit block comprising the ground truth representation of a second circuit block and wherein the circuit block simulator module is configured to perform analog defect simulation (ADS) based on defects injected into the ground truth circuit block.

20. The simulator module of claim 19 , further comprising a test coverage metric determination module configured to determine an original test coverage metric of a test program by executing the test program based on an ADS of a circuit under test (CUT) comprising the FAAM circuit block and the ground truth circuit block, wherein the original test coverage defines a probability that the test program fails when there is a defect in an original circuit associated with the CUT, wherein the original circuit corresponds to a circuit when each circuit block within the CUT is implemented using a corresponding ground truth representation.

21. The simulator module of claim 20 , wherein the test coverage metric determination module is configured to determine the original test coverage metric, in accordance with a predefined original test coverage metric relation that is derived based on a first probability function that relates an output of the original circuit to the output of the CUT and a second probability function comprising a probability density of the output of the CUT, given there is a defect in the FAAM.

22. The simulator module of claim 20 , wherein the test coverage metric determination module is further configured to determine a FAAM test coverage metric of the test program by executing the test program based on the ADS performed on the CUT, wherein the FAAM test coverage metric defines a probability that the test program fails when there is a defect in the CUT.

23. The simulator module of claim 18 , further comprising an application program interface (API) module configured to couple the circuit block simulator module to one or more machine learning libraries, in order to enable the circuit block simulator module to simulate the FAAM circuit block.

24. A simulator module, comprising: a circuit block simulator module configured to simulate a behavioral model of a first circuit block, wherein the behavioral model comprises a computable parametric transformation that maps elements from an input set to elements from an output set associated with the first circuit block and wherein the behavioral model mimics a functionality of a ground truth of the first circuit block; and an application program interface (API) module configured to couple the circuit block simulator module to one or more machine learning libraries, in order to enable the circuit block simulator module to simulate the behavioral model, when the behavioral model is compatible with the one or more machine learning libraries.

25. The simulator module of claim 24 , wherein the circuit block simulator module is further configured to simulate a ground truth circuit block comprising a ground truth representation of a second circuit block and wherein the circuit block simulator module is configured to perform analog defect simulation (ADS) based on defects injected into the ground truth circuit block.